Hi Jason, 



Could you zip/tar an example dataset or two and upload here: https://oxfile.ox.ac.uk/oxfile/work/extBox?id=820635FD10D06DE486



Also provide any masks you are using and featquery commands (if command line). 





Cheers,



Taylor Hanayik





Taylor Hanayik PhD

Analysis Research Software Engineer

FMRIB, John Radcliffe Hospital

University of Oxford

[log in to unmask]











> On 18 Dec 2019, at 21:07, Jason Oliver, Ph.D. <[log in to unmask]> wrote:

> 

> Hi everyone,

> I am writing in the hopes that someone out there can clarify some seemingly unusual results we are getting from featquery. I am extracting a series of COPEs at both the first level (individual runs) and second level (a single fixed effect that just collapses across runs).

>  

> The “number of voxels” value varies significantly across subjects and even across different runs for the same subject at the first level. The report correctly identifies the mask as being in standard space, so I am unsure why the number of voxels would differ as presumably everything is being warped to standard space prior to the extraction. At the second level, the number of voxels is constant across subjects though an order of magnitude larger (i.e. my amygdala ROI is typically 15-25 voxels at the first level and 240 at the second level). The second level report indicates the mask is in space of example_func, though unless I am missing something example_func should also be standard space at this stage of the analysis.

>  

> A few other notes:

> -          The masks were based off the MNI template and I am certain are in standard space.

> -          Registration appears to be working correctly for subjects (no errors in the report)

> -          Despite the above, attempts to load individual COPEs from the stats folder at the first level over the MNI template DO generate an error about incompatible overlays. Yet it works fine at the second level. These are clearly being retained in different dimensions despite the registration working. I <assume> this relates to the issue at hand, but do not know if featquery could be using the transformation matrix to convert these to standard space

> -          Correlation of COPE values at the second level with the average of the 4 first level runs is relatively high (~.85). I assume we wouldn’t expect a perfect correlation here because of potential differences in how percent signal chance is being computed. Yet if featquery was using the same coordinate values from wildly different coordinate spaces, I would not expect the correlation to be nearly that high.

>  

> Appreciate any input anyone can provide! We are stumped here and I am extremely hesitant to move forward until I know what is going on with this…

>  

> Best,

> Jason

>  

> Jason A. Oliver, Ph.D.

> Assistant Professor

> Department of Psychiatry and Behavioral Sciences

> Duke University School of Medicine

> Office Phone: 919-668-0093

> Email Address: [log in to unmask]

>  

> 

> To unsubscribe from the FSL list, click the following link:

> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

> 





########################################################################



To unsubscribe from the FSL list, click the following link:

https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1